Comparison of Human and Drosophila Atlastin Gtpases

Protein Cell 2015, 6(2):139–146 DOI 10.1007/s13238-014-0118-0 Protein & Cell

RESEARCH ARTICLE Comparison of human and Drosophila atlastin GTPases

Fuyun Wu1,2, Xiaoyu Hu1,2, Xin Bian1,2, Xinqi Liu3,4, Junjie Hu1,2,5&

1 Department of Genetics and Cell Biology, College of Life Sciences, Nankai University, Tianjin 300071, China 2 Tianjin Key Laboratory of Protein Sciences, Tianjin 300071, China 3 Department of Biochemistry and Molecular Biology, College of Life Sciences, Nankai University, Tianjin 300071, China 4 State Key Laboratory of Medicinal Chemical Biology, Tianjin 300071, China 5 National Laboratory of Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, China & Correspondence: [email protected] (J. Hu)

Received October 9, 2014 Accepted October 27, 2014 Cell &

ABSTRACT KEYWORDS endoplasmic reticulum, membrane fusion, atlastin, GTPase, X-ray crystallography Formation of the endoplasmic reticulum (ER) network

requires homotypic membrane fusion, which involves a Protein class of atlastin (ATL) GTPases. Purified Drosophila ATL INTRODUCTION is capable of mediating vesicle fusion in vitro, but such activity has not been reported for any other ATLs. Here, In eukaryotic cells, the endoplasmic reticulum (ER) is com- we determined the preliminary crystal structure of the posed of interconnected tubules and sheets (Shibata et al., cytosolic segment of Drosophila ATL in a GDP-bound 2006). The merging of ER membranes, i.e. homotypic state. The structure reveals a GTPase domain dimer with fusion, is required for ER network formation and mediated by the subsequent three-helix bundles associating with a dynamin-like membrane bound GTPase family known as their own GTPase domains and pointing in opposite atlastins (ATLs) (Hu et al., 2009; Orso et al., 2009). No ATL directions. This conformation is similar to that of human has been found in yeast and plant cells, but a similar class of ATL1, to which GDP and high concentrations of inor- GTPases, named Sey1p or RHD3, plays an analogous role ganic phosphate, but not GDP only, were included. in ER morphogenesis (Hu et al., 2009; Anwar et al., 2012; Drosophila ATL restored ER morphology defects in Zhang et al., 2013). Depletion of Drosophila or zebrafish ATL mammalian cells lacking ATLs, and measurements of results in neuronal defects (Lee et al., 2008; Lee et al., 2009; nucleotide-dependent dimerization and GTPase activity Fassier et al., 2010). Deletion or mutation of RHD3 leads to were comparable for Drosophila ATL and human ATL1. defects in plant growth and short and wavy root hairs However, purified and reconstituted human ATL1 (Schiefelbein and Somerville, 1990; Wang et al., 1997; exhibited no in vitro fusion activity. When the cytosolic Stefano et al., 2012). Depletion or mutation of human ATL1, segment of human ATL1 was connected to the trans- a dominant form in the central nervous system, causes membrane (TM) region and C-terminal tail (CT) of Dro- unbranched ER at the cellular level (Hu et al., 2009) and sophila ATL, the chimera still exhibited no fusion hereditary spastic paraplegia (HSP), a neurodegenerative activity, though its GTPase activity was normal. These disease characterized by axon shortening in corticospinal results suggest that GDP-bound ATLs may adopt mul- motor neurons and progressive spasticity and weakness of tiple conformations and the in vitro fusion activity of ATL the lower limbs, at the organism level (Zhao et al., 2001; cannot be achieved by a simple collection of functional Salinas et al., 2008). domains. Several assays have been developed to study the fusogenic activity of these GTPases. First, yeast cells lacking Sey1p and either Yop1p or Rtn1p, members of the ER tubule- forming protein family, exhibit abnormal ER morphology (Hu et al., 2009). Restoration of the ER network can then be Fuyuan Wu and Xiaoyu Hu contributed equally to this work. monitored upon expression of any ER fusogen candidates

A GTPase 3HBTMsCT C dmATL GTPase-1 GTPase-2 hsATL1

B Form 1 Form 2 Form 3

3HB-1 3HB-2 +GDP

+ - +GDP +GDP/Pi GDP/AlF 4 +GppNp D E

hsATL1 + GDP/Pi hsATL1 + GDP/Pi Cell

& dmATL + GDP dmATL + GDP

Figure 1. Crystal structure of the cytosolic domain of Drosophila ATL. (A) Schematic diagrams of Drosophila ATL and human ATL1. (B) Human ATL1 dimers observed in crystal form 1 (PDB code: 3QNU, left, GDP-bound), form 2 (PDB code: 3QOF, middle,

Protein GDP-bound in the presence of inorganic phosphate), and form 3 (PDB code: 4IDO, right, GDP/AlF4-bound or GppNp-bound) are shown. The protomers in the dimer are shown in green and purple cartoon representation. (C) Structure of the GDP-bound form of Drosophila ATL. The protomers in the dimer are shown in green and purple cartoon representation. GDP is shown in orange stick representation, and magnesium ion is shown as a yellow sphere. 3HB, three-helix bundle. (D) Structural comparison of human and Drosophila ATLs. Human ATL1 (hsATL1) is shown in pink and Drosophila ATL (dmATL) is shown in green. GDP in dmATL is shown in orange, and magnesium ion is shown in yellow. (E) As in C, but with a dimer. Superposition was performed using the protomers on the left.

(Anwar et al., 2012; Zhang et al., 2013). Second, ER fusion facing each other, but the position of the 3HB differs can be followed during mating of haploid yeast cells carrying (Fig. 1B). In form 1, ATL1 is in complex with GDP, the two cytosolic and ER markers (Anwar et al., 2012). Deletion of 3HBs associate with the paired GTPase domains and form a Sey1p significantly reduces the fusion process, whereas crossover conformation. In this case, the connecting TMs expression of a functional ER fusogen, such as human ATL1 would have to sit in the same membranes; thus, the structure or RHD3 (Anwar et al., 2012; Zhang et al., 2013), can rescue corresponds to a “post-fusion” state in which the membranes this defect. Moreover, purified drosophila ATL has been have already fused. In form 2, ATL1 is also GDP-bound, but reconstituted into proteoliposomes, and signals of lipid mixing in the presence of high concentration of inorganic phos- in the presence of magnesium and GTP provide direct evi- phate. The two 3HBs associate with their own GTPase dence for membrane fusion mediated by ATL (Orso et al., domains and point in opposite directions. This structure 2009; Bian et al., 2011). Using this assay, Sey1p and mem- implies that two ATL molecules likely sit in apposing members of the RHD3 family have been confirmed to mediate branes, corresponding to a membrane-tethered “pre-fusion”

fusion in vitro. However, drosophila ATL is the only ATL that state. In form 3, ATL is crystallized with either GDP/AlF4 or has been reported to have fusogenic activity in vitro. GppNp. The two 3HBs come even closer than in form 1. ATL contains an N-terminal GTPase and a three-helix Collectively, ATL-mediated fusion requires dimerization bundle (3HB), followed by two closely spaced transmem- resulting from GTP binding and conformational changes brane (TM) segments and a C-terminal tail (CT) (Fig. 1A). In induced by GTP hydrolysis (Bian et al., 2011; Byrnes and an effort to understand the ATL-associated fusion mecha- Sondermann, 2011; Hu et al., 2011; Lin et al., 2012) nisms, three crystal structures of the N-terminal cytosolic In addition to the N-terminal cytosolic domain, the TM and domain of human ATL1 have been determined (Bian et al., CT have also been shown to play important roles in fusion 2011; Byrnes and Sondermann, 2011). In all three forms of (Liu et al., 2012). The TM of ATL appears to be more than a ATL, the molecule forms a dimer with the GTPase domains membrane anchor, as deletion or replacement of the TM

140 © The Author(s) 2014. This article is published with open access at Springerlink.com and journal.hep.com.cn Comparing human and Drosophila ATL RESEARCH ARTICLE

region results in a loss of fusion activity. The CT of ATL forms Table 1. Data collection and refinement statistics an amphipathic helix that binds and destabilizes the membranes to facilitate fusion. A synthetic peptide corresponding Data collection to parts of the CT of Drosophila ATL was able to restore Space group C121 fusion defects in the tailless ATL mutant. The importance of Wavelength (Å) 1.54 ATL-CT is consistent with the fact that a lack of CT in human Unit cell ATL1 mutants causes HSP (Liu et al., 2012). Drosophila ATL has been used extensively to study a, b, c (Å) 71.62, 63.91, 107.76 membrane fusion in vitro, but structural information about α, β, γ (°) 90, 101.6, 90 ATL has been obtained purely from human ATL1. Whether Molecules/ASU 1 human ATL can mediate fusion in vitro is also not clear. Resolution range (Å)a 50–2.87 (2.92–2.87) Here, we determined the structure of the N-terminal cytosolic a domain of Drosophila ATL and compared fusogenic features Completeness (%) 91.93 (80.49) of human and Drosophila ATL. Redundancya 3.1 (2.5) No. of total reflections 31354 No. of unique reflections 10190 RESULTS I/σa 26.4 (3.8) Crystal structure of the cytosolic domain of Drosophila Rsym (%)a,b 8.1 (31.5) ATL Refinement statistics Cell

To gain insight into Drosophila ATL-mediated membrane Resolution (Å) 2.87 & fusion, we determined the crystal structure of cytATL. Resi- No. of reflections 10185 dues 1–422 were expressed in E. coli, purified, and crys- c tallized in the presence of GDP. A centered monoclinic Rwork/Rfree (%) 27.30/36.40 crystal form was used to determine the structure at 2.87-Å No. of atoms resolution by molecular replacement using the human ATL1 Protein 2928 Protein structure (PDB code: 3QOF) as a search model (Table 1). Ligands 29 Despite numerous trials for crystal optimization, we were unable to improve the data quality; the final resolution is Water 6 acceptable for phase determination and preliminary struc- B-factors (Å2) tural interpretation. One ATL molecule was present in the Protein 41.49 asymmetric unit, but a crystallographic dimer was observed Water 41.64 along one of the two-fold symmetry axes. Ligands 28.41 Similar to human ATL1, the structure of Drosophila cytATL revealed a GTPase domain and a 3HB connected by a linker r.m.s.d. region (Fig. 1C). In the GDP-bound dimer, the GTPase Bond length (Å) 0.009 domains interact with each other. The two 3HBs dock to their Bond angle 1.487 own GTPase domains and point in opposite directions. This Ramachandran analysis configuration is very similar to form 2, or the “pre-fusion” state, of human ATL1 (Fig. 1B). In fact, the cytATLs of human Most favored (%) 65.0 and Drosophila are virtually superimposable (Fig. 1D). Only Additional allowed (%) 27.9 part of the 3HB showed appreciable differences between the Generously allowed (%) 4.7 two structures (Fig. 1D). When the dimers of human and Disallowed (%) 2.4 Drosophila ATL were compared, the relative position a between two protomers was slightly different (Fig. 1E). Values in parentheses are for the highest resolution shell. b Notably, the dimer of Drosophila ATL was observed with Rsym = Σ|I − |/Σ, where I is the observed intensity, and fl GDP, but the similar dimer of human ATL1 was seen with is the average intensity of multiple symmetry-related re ection GDP and a high concentration of inorganic phosphate. When observations. c Σ − Σ only GDP was added, a completely different human ATL1 R = hkl|Fobs| |Fcalc|/ hkl|Fobs|. Rfree was calculated from 5% of fl fi dimer (form 1 in Fig. 1B) was formed. In addition, when the re ections excluded from re nement. average B-factors were calculated for the GTPase domain and the 3HB in our structure (37.84 for GTPase and 60.59 The active site of Drosophila ATL is mostly very similar to for 3HB), the results suggested that the 3HB exhibited sig- that of human ATL1 (Fig. 2), where conserved residues from nificantly higher flexibility. These results suggest that the four signature motifs engage the nucleotide in the presence 3HB of ATL may adopt more than one conformation upon of magnesium ion. Notably, R48 of Drosophila ATL points GDP binding. towards GDP, while the equivalent Arginine in human ATL1

© The Author(s) 2014. This article is published with open access at Springerlink.com and journal.hep.com.cn 141 RESEARCH ARTICLE Fuyun Wu et al.

hs-GDP hs-GDP

dm-GDP dm-GDP dm-S52 dm-S52

dm-D121hs-S81 dm-D121 hs-S81 hs-R77 hs-R77 dm-R48 dm-R48 Mg2+ Mg2 dm-K51 dm-K51 hs-K80 hs-K80 hs-D146 hs-D146

Figure 2. Comparison of the active site. Stereoview of the superimposed active sites of Drosophila ATL (colored as in Fig. 1C) and human ATL1 (in pink, PDB ID code 3QOF). Critical residues are highlighted. The nucleotides are shown in stick representation and the magnesium ions as spheres.

Cell (R77) points away from GDP, and forms a salt bridge with a membranes, and GTP hydrolysis drives conformational

& Glutamic acid in the pairing GTPase (Bian et al., 2011). In changes in ATL. To test whether Drosophila and human the recent human ATL1 structures, where GppNp or GDP/ ATLs differ in dimerization tendencies upon GTP binding, we

AlF4 were added, R77 points at the nucleotides (Byrnes compared the dimerization state of puriﬁed cytATLs using et al., 2013). Thus, this Arginine was proposed to ﬂip analytic ultracentrifugation (AUC). As shown previously between these conformations during GTP hydrolysis: it (Bian et al., 2011), but to a lesser extent, human cytATL1

Protein might facilitate GTPase dimer formation in the outward formed some dimer in the presence of GppNp, a non-hy- position and act as an Arg-finger in the inward position. Even drolyzable analog of GTP, at a concentration of 0.03 mmol/L though the electron density for R48 is poor in our structure, it (Fig. 4A). A similar tendency of dimerization was observed is consistent that this critical arginine has the flexibility to when Drosophila cytATL was tested in the presence of switch between conformations, especially in the GDP-bound GppNp (Fig. 4B). In contrast, little dimer was seen when both state. proteins were tested at the same concentrations but with GDP (Fig. 4A and 4B). We then compared the GTPase activities of purified cytATLs. When the GTPase assay was Drosophila ATL can functionally replace human ATLs performed using Drosophila or human cytATLs at a con- To analyze the function of Drosophila ATL, we tested whe- centration of 1 μmol/L, similar rates of GTP hydrolysis were ther it can replace ATLs in mammalian cells. We first used detected (Fig. 4C, left panel). The same increase in hydro- small interference RNA (siRNA) to deplete major ATL pro- lysis rates was measured when higher concentrations teins in COS-7 cells and visualized the ER morphology using (2 μmol/L and 5 μmol/L) of the proteins were tested (Fig. 4C, calreticulin as a marker. The COS-7 cell line expresses little middle and right panels). Taken together, these results ATL1 but sufficient ATL2 and ATL3. Consistent with previous suggest that Drosophila and human ATLs display the same results, when ATL2 and ATL3 were partially knocked down in tendencies in nucleotide-dependent dimerization and the COS-7 cells, the ER in most of the cells exhibited unbran- same GTP hydrolysis ability. ched morphology (Fig. 3A, 3B, and 3D), indicating defects in ER fusion. As expected, expression of human ATL1 effi- Testing human and Drosophila ATLs in a fusion assay ciently restored the ER network in double-depleted cells (Fig. 3A, 3C, and 3D). When Drosophila ATL was expressed To compare the membrane fusion activity of full-length ATLs, at a comparable level, significantly less cells exhibited we used a previously established in vitro lipid-mixing assay. unbranched ER phenotype (Fig. 3A, 3C, and 3D). These Drosophila and human ATLs were purified and reconstituted results indicated that Drosophila ATL and human ATL1 are at equal concentrations into donor and acceptor proteolipo- equally active in maintaining the ER network. somes (Fig. 5A). The donor vesicles contained lipids labeled with two fluorophores (NBD and rhodamine) at quenching concentrations; the fusion with unlabeled acceptor vesicles Dimer formation and GTP hydrolysis of Drosophila resulted in dilution and dequenching of NBD. Drosophila ATL and human ATLs resulted in efficient fusion in the presence of GTP and Mg2+ ATL-mediated fusion has been shown to require GTP- (Fig. 5B) as reported previously (Bian et al., 2011). Consis- dependent dimerization between ATL molecules in apposing tently, no fusion was seen with GDP,GppNp, or the absence of

142 © The Author(s) 2014. This article is published with open access at Springerlink.com and journal.hep.com.cn Comparing human and Drosophila ATL RESEARCH ARTICLE

A C Flag Calreticulin Merge Flag-hsATL1 Flag-dmATL Flag-hsATL1 Control siRNA ATL2&3 siRNA ATL2&3 siRNA ATL2&3 IB: ATL2 * ATL2 siRNA ATL2&3

IB: ATL3 ATL3 Flag Calreticulin Merge hsATL1 IB: Flag dmATL

IB: GAPDH GAPDH Flag-dmATL siRNA ATL2&3

B D Normal ER Calreticulin Unbranched ER 100

60 Cell

40 & Calreticulin ER morphology 20

% Of cells with indicated 0 Protein siRNA Control dmATL hsATL1 siRNA siRNA ATL2&3 ATL2&3 ATL2&3 siRNA ATL2&3 Control

Figure 3. Functional tests of Drosophila ATL in mammalian cells. (A) COS-7 cells were transfected with siRNA oligonucleotides as indicated. Levels of endogenous ATL2 and ATL3 or exogenously expressed Drosophila ATL and human ATL1 were determined by immunoblotting. GAPDH was used as a loading control. Asterisk (*) indicates a non-speciﬁc band. (B) The ER morphology of COS-7 cells was visualized using calreticulin, an endogenous luminal ER protein, and indirect immunoﬂuorescence using a confocal microscope. Scale bar = 10 μm. (C) As in (B), but with double-depleted COS-7 cells transfected with Flag-human ATL1, or Drosophila ATL. (D) The ER morphology of samples shown in (B) and (C) was categorized as “normal” or “unbranched”. A total of 80–150 cells were counted for each sample. All graphs are representative of three repetitions.

Mg2+ (Fig. 5B). Surprisingly, human ATL1 was inactive under because fruit flies serve as a great genetic model for probing the same conditions, even with higher protein concentrations the physiological role of ER fusion, but also because it in the proteoliposomes compared to Drosophila ATL (Fig. 5B). enables a powerful system for studying homotypic mem- Because the human and Drosophila cytATLs exhibit brane fusion in vitro. Our results here provide new insights equivalent GTPase activity, the TM and CTof Drosophila may into ATL-mediated fusion and clarify a few possibilities for convey more efficient membrane fusion in vitro. To test this further understanding this process. hypothesis, we replaced the TM-CT of human ATL1 with that The crystal structure of Drosophila ATL in complex with of Drosophila ATL and measured the GTPase activity and GDP forms an unexpected second dimer form similar to membrane fusion of the chimera. As expected, the detergent- human ATL1 in the “pre-fusion” state. This observation solubilized and purified ATL chimera hydrolyzed GTP at the does not necessarily mean that ATL will switch from “post- same rates as Drosophila and human ATL (Fig. 5C). How- fusion” to “pre-fusion” directly in the presence of GDP, but ever, membrane fusion was still not detected (Fig. 5B). These rather raises the possibility that GDP-bound ATL is simply results suggest that fusion mediated by ATLs requires coor- flexible and may adopt multiple conformations before dination between different domains of the molecule. releasing GDP and recycling GTPase. Previous results indicated that ATL remains a relatively loose dimer in the presence of GDP (Bian et al., 2011). Our new data with DISCUSSION both human and Drosophila cytATLs tested at a lower Drosophila ATL has played important roles in understanding concentration clearly showed that GDP-bound ATL is homotypic ER fusion mediated by the ATL family, not only mostly monomeric. These results favor the notion that the

© The Author(s) 2014. This article is published with open access at Springerlink.com and journal.hep.com.cn 143 RESEARCH ARTICLE Fuyun Wu et al.

A B 20 40 42.7/51.3/40.3 42/49.5/35.1 dmATL 15 hsATL1 30 hsATL1 dmATL + GDP + GDP 10 20

c (s) dmATL hsATL1 c (s) + + GppNp GppNp 5 10 72.5 71.5 0 0 012345 012345 Sed coefficient [S] Sed coefficient [S] C 0.25 1 μmol/L 0.25 2 μmol/L 0.25 5 μmol/L hsATL1 dmATL 0.2 0.2 0.2 Control

360 0.15 360 0.15 360 0.15 0.1 0.1 0.1 OD 0.05 OD 0.05 OD 0.05 0 0 0 0 10 20 30 0 10 20 30 0 10 20 30 Time (min) Time (min) Time (min)

Figure 4. Biochemical comparison of cytATLs. (A) Nucleotide-dependent dimerization of the N-terminal cytosolic domain of Cell hsATL1 was determined at a concentration of 0.03 mmol/L by analytical ultracentrifugation in the presence of the indicated

& nucleotides. (B) As in (A), but with dmATL. (C) The GTPase activities of 1 μmol/L, 2 μmol/L, and 5 μmol/L of N-terminal cytosolic domains of human ATL1 (hsATL1) and Drosophila ATL (dmATL) were measured by phosphate release at saturating concentrations of GTP (0.5 mmol/L). All graphs are representative of three repetitions. Protein transition from dimer to monomer after GTP hydrolysis will tends to form puncta around the junctions of the tubular ER occur spontaneously. network (Moss et al., 2011), but human ATL1 is evenly dis- Drosophila ATL and human ATL1 have very high tributed along ER membranes. Whether these differences sequence similarity (Bian et al., 2011). Almost all key resi- are linked to the different in vitro activities of the two ATLs dues identified so far in the ATL family can be found in both remains to be investigated. ATLs. Thus, human ATL1 is expected to exhibit fusion activity in vitro, just like Drosophila ATL. Surprisingly, our MATERIALS AND METHODS results reveal that human ATL1 is inactive, even under the same conditions as Drosophila ATL. Structural and bio- Protein expression, purification, and crystallization chemical analyses revealed that GTPase activity is critical The cytosolic domain of human ATL-1 (residues 18–447) and full- for the fusion reaction. In addition, the TM and CT regions of length Drosophila ATL were expressed and purified as described the molecule play important roles. However, none of these previously (Bian et al., 2011). The cytosolic domain of Drosophila factors seem to be the determinants of the in vitro fusogenic ATL (cytATL; residues 1–413) was cloned into the pET-28a vector, activity of human ATL1. First, human ATL1 has same rate of expressed, and purified in the same manner as human ATL1. The GTP hydrolysis as Drosophila ATL, which is active in in vitro full-length human ATL1 and chimeric ATL were amplified and cloned fusion. Second, the TM-CT of Drosophila ATL could not help into the pGEX-6P-1 vector and expressed in Escherichia coli as human ATL1 with its fusion reaction when the active N-ter- GST fusion proteins. The proteins were purified in the same manner minal cytosolic domain of human ATL1 is attached. Based on as full-length Drosophila ATL, and the purified proteins were con- the depletion phenotype and the results of the yeast in vivo centrated to 1 mg/mL for subsequent analysis. fusion assay (Hu et al., 2009; Anwar et al., 2012), human Drosophila cytATL protein (6 mg/mL) was incubated with 1 mmol/L ATL1 is clearly an active ER fusogen in vivo. Unfortunately, it GDP for 1 h at 4°C, and then crystallized using the hanging drop appears that the ATL activity in the in vitro fusion assays vapor diffusion method equilibrated against a reservoir solution of cannot be predicted based on its in vivo behavior. 22% PEG3350, 0.1 mol/L Tris, 0.2 mol/L NaCl, pH 8.5. Crystals were The lack of in vitro fusion activity for human ATL1 sug- grown for one week at 20°C. gests that it may be a less active fusion enzyme than Dro- sophila ATL, but its activity is elevated by unclear Data collection and structure determination mechanisms in vivo. These two ATLs behave differently in cells in at least two respects. First, depletion of Drosophila Diffraction data for the crystal were collected at 100 K using a ATL causes ER fragmentation (Orso et al., 2009), but wavelength of 1.54 Å at a home-source X-ray facility (Rigaku Mi- depletion of human ATL does not. Second, Drosophila ATL croMax007HF) at Nankai University. For cryo-protection, the crystals

144 © The Author(s) 2014. This article is published with open access at Springerlink.com and journal.hep.com.cn Comparing human and Drosophila ATL RESEARCH ARTICLE

A Top Bottom Top Bottom Top Bottom 130 130 130 100 100 100 70 70 70 55 * 55 * 55 * 40 35 35 35 25 25 25 M67891012345 M1112131415M hsATL1 dmATL Chimeric ATL B 25 dmATL 20 hsATL1 15 Chimeric ATL 10 GDP 5 GMppNp W/O Mg2+ 0

% Total fluorescence% Total -5 5101520 25 30 Time (min)

C 1 μmol/L 2 μmol/L 5 μmol/L 0.16 0.16 0.16 0.12 0.12 0.12 hsATL1 Cell 360 360 360 dmATL 0.08 0.08 0.08

Chimeric ATL & OD OD OD 0.04 0.04 0.04 Control 0 0 0 0 10 20 30 0 10 20 30 0 10 20 30 Time (min) Time (min) Time (min)

Figure 5. Comparison of membrane fusion in vitro. (A) Full-length ATLs were reconstituted into proteoliposomes. Flotation in a Protein sucrose gradient indicated efﬁcient reconstitution of the proteins. (B) Full-length ATLs were reconstituted at equal concentrations into donor and acceptor vesicles. The lipid to protein ratio used for each protein is indicated. Fusion was monitored by dequenching the NBD-labeled lipids present in the donor vesicles and was initiated by the addition of GTP. Control experiments were performed using chimeric ATL in the absence of Mg2+ or the presence of GDP or GppNp instead of GTP. (C) The GTPase activities of puriﬁed full-length ATLs (1 μmol/L, 2 μmol/L, and 5 μmol/L) were measured by phosphate release at saturating concentrations of GTP (0.5 mmol/L). All graphs are representative of three repetitions.

were soaked in mother liquor plus 20% (v/v) glycerol for a few Flotation assay seconds before data collection. All data sets were processed with The reconstitution efficiency was determined by flotation of proteo- HKL2000 and converted to CCP4 format. The structure was deter- liposomes on a sucrose gradient. Proteoliposomes (30 µL) were mined by molecular replacement using the human ATL1 structure mixed with 100 µL of 1.9 mol/L sucrose and overlaid with 100 µL of (Protein Data Bank code 3QOF) as a search model. The preliminary 1.25 mmol/L sucrose and 20 µL of 0.25 mmol/L sucrose, all in models were built and refined with the Phenix program. 25 mmol/L HEPES (pH 7.5). The samples were centrifuged in a Beckman TLS 55 rotor at 55,000 rpm and 4°C for 75 min. The gradient was fractionated into five 50-µL fractions and analyzed by GTPase activity assay SDS-PAGE stained with Coomassie. GTPase activity was determined using the EnzChek phosphate assay kit (Invitrogen). Reactions were performed in a 100 µL volume Fusion assay and initiated by the addition of 0.5 mmol/L GTP. Absorbance at The in vitro lipid-mixing assay was performed as described previ- 360 nm was measured at 1-min intervals for 30 min at 37°C. ously (Bian et al., 2011). In brief, the full-length ATL proteins were reconstituted into the liposome, and the labeled donor proteoliposomes were mixed with unlabeled acceptor proteoliposomes in the Analytical ultracentrifugation presence of 5 mmol/L Mg2+ in a total volume of 100 µL per reaction. Sedimentation velocity data were collected at a speed of 142,249 ×g The fusion reaction was started by the addition of 0.5 mmol/L GTP in an An-60 Ti rotor at 4°C. The proteins were prepared in 50 mmol/L and NBD fluorescence measured at 1-min intervals at 37°C. After

Tris (pH 8.0), 5 mmol/L MgCl2, 1 mmol/L DTT, and 1 mmol/L 30 min, 5 µL of 10% (w/v) DDM was added to determine the total nucleotide. Data were analyzed by the program SEDFIT NBD ﬂuorescence. Fusion was expressed as the percentage of total (Version 11.8). ﬂuorescence.

© The Author(s) 2014. This article is published with open access at Springerlink.com and journal.hep.com.cn 145 RESEARCH ARTICLE Fuyun Wu et al.

ACCESSION CODE Lee Y, Paik D, Bang S, Kang J, Chun B, Lee S, Bae E, Chung J, Kim J (2008) Loss of spastic paraplegia gene atlastin induces age- Coordinates and structure factors have been deposited with dependent death of dopaminergic neurons in Drosophila. Neu- RCSB accession code: 3X1D. robiol Aging 29:84–94 Lee M, Paik SK, Lee MJ, Kim YJ, Kim S, Nahm M, Oh SJ, Kim HM, ACKNOWLEDGEMENTS Yim J, Lee CJ et al (2009) Drosophila atlastin regulates the This work was supported by the National Natural Science Founda- stability of muscle microtubules and is required for synapse – tion of China (Grant No. 31225006) and an International Early development. Dev Biol 330:250 262 Career Scientist grant from the Howard Hughes Medical Institute. Lin S, Sun S, Hu J (2012) Molecular basis for sculpting the endoplasmic reticulum membrane. Int J Biochem Cell Biol 44:1436–1443 COMPLIANCE WITH ETHICS GUIDELINES Liu TY, Bian X, Sun S, Hu X, Klemm RW, Prinz WA, Rapoport TA, Hu Fuyun Wu, Xiaoyu Hu, Xin Bian, Xinqi Liu, and Junjie Hu declare J (2012) Lipid interaction of the C terminus and association of the that they have no conflict of interest. transmembrane segments facilitate atlastin-mediated homotypic – This article does not contain any studies with human or animal endoplasmic reticulum fusion. Proc Natl Acad Sci 109:E2146 subjects performed by the any of the authors. E2154 Moss TJ, Andreazza C, Verma A, Daga A, McNew JA (2011) OPEN ACCESS Membrane fusion by the GTPase atlastin requires a conserved C-terminal cytoplasmic tail and dimerization through the middle This article is distributed under the terms of the Creative Commons domain. Proc Natl Acad Sci 108:11133–11138 Cell Attribution License which permits any use, distribution, and Orso G, Pendin D, Liu S, Tosetto J, Moss TJ, Faust JE, Micaroni M, & reproduction in any medium, provided the original author(s) and Egorova A, Martinuzzi A, McNew JA et al (2009) Homotypic the source are credited. fusion of ER membranes requires the dynamin-like GTPase atlastin. Nature 460:978–983 Salinas S, Proukakis C, Crosby A, Warner TT (2008) Hereditary REFERENCES spastic paraplegia: clinical features and pathogenetic mecha- Protein nisms. Lancet Neurol 7:1127–1138 Anwar K, Klemm RW, Condon A, Severin KN, Zhang M, Ghirlando Schiefelbein JW, Somerville C (1990) Genetic control of root hair R, Hu J, Rapoport TA, Prinz WA (2012) The dynamin-like development in Arabidopsis thaliana. Plant Cell 2:235–243 GTPase Sey1p mediates homotypic ER fusion in S. cerevisiae. Shibata Y, Voeltz GK, Rapoport TA (2006) Rough sheets and J Cell Biol 197:209–217 smooth tubules. Cell 126:435–439 Bian X, Klemm RW, Liu TY, Zhang M, Sun S, Sui X, Liu X, Rapoport Stefano G, Renna L, Moss T, McNew JA, Brandizzi F (2012) In TA, Hu J (2011) Structures of the atlastin GTPase provide insight Arabidopsis, the spatial and dynamic organization of the endo- into homotypic fusion of endoplasmic reticulum membranes. Proc fl Natl Acad Sci 108:3976–3981 plasmic reticulum and Golgi apparatus is in uenced by the Byrnes LJ, Sondermann H (2011) Structural basis for the nucleotide- integrity of the C-terminal domain of RHD3, a non-essential – dependent dimerization of the large G protein atlastin-1/SPG3A. GTPase. Plant J 69:957 966 Proc Natl Acad Sci 108:2216–2221 Wang H, Lockwood SK, Hoeltzel MF, Schiefelbein JW (1997) The ROOT HAIR DEFECTIVE3 gene encodes an evolutionarily Byrnes LJ, Singh A, Szeto K, Benvin NM, O’Donnell JP, Zipfel WR, conserved protein with GTP-binding motifs and is required for Sondermann H (2013) Structural basis for conformational switch- regulated cell enlargement in Arabidopsis. Genes Dev 11:799– ing and GTP loading of the large G protein atlastin. EMBO J 32 811 (3):369–384 Zhang M, Wu F, Shi J, Zhu Y, Zhu Z, Gong Q, Hu J (2013) ROOT Fassier C, Hutt JA, Scholpp S, Lumsden A, Giros B, Nothias F, HAIR DEFECTIVE3 family of dynamin-like GTPases mediates Schneider-Maunoury S, Houart C, Hazan J (2010) Zebrafish homotypic endoplasmic reticulum fusion and is essential for atlastin controls motility and spinal motor axon architecture via – inhibition of the BMP pathway. Nat Neurosci 13:1380–1387 Arabidopsis development. Plant Physiol 163:713 720 Hu J, Shibata Y, Zhu P-P, Voss C, Rismanchi N, Prinz WA, Rapoport Zhao X, Alvarado D, Rainier S, Lemons R, Hedera P, Weber CH, TA, Blackstone C (2009) A class of dynamin-like GTPases Tukel T, Apak M, Heiman-Patterson T, Ming L et al (2001) fi involved in the generation of the tubular ER network. Cell Mutations in a newly identi ed GTPase gene cause autosomal dominant hereditary spastic paraplegia. Nat Genet 29:326–331 138:549–561 Hu J, Prinz WA, Rapoport TA (2011) Weaving the web of ER tubules. Cell 147:1226–1231